Research Article www.acsami.org
Interactive Effect of Hysteresis and Surface Chemistry on Gated Silicon Nanowire Gas Sensors Yair Paska and Hossam Haick* The Department of Chemical Engineering and Russell Berrie Nanotechnology Institute, TechnionIsrael Institute of Technology, Haifa 32000, Israel S Supporting Information *
ABSTRACT: Gated silicon nanowire gas sensors have emerged as promising devices for chemical and biological sensing applications. Nevertheless, the performance of these devices is usually accompanied by a “hysteresis” phenomenon that limits their performance under realworld conditions. In this paper, we use a series of systematically changed trichlorosilane-based organic monolayers to study the interactive effect of hysteresis and surface chemistry on gated silicon nanowire gas sensors. The results show that the density of the exposed or unpassivated Si−OH groups (trap states) on the silicon nanowire surface play by far a crucial effect on the hysteresis characteristics of the gated silicon nanowire sensors, relative to the effect of hydrophobicity or molecular density of the organic monolayer. Based on these findings, we provide a tentative model-based understanding of (i) the relation between the adsorbed organic molecules, the hysteresis, and the related fundamental parameters of gated silicon nanowire characteristics and of (ii) the relation between the hysteresis drift and possible screening effect on gated silicon nanowire gas sensors upon exposure to different analytes at real-world conditions. The findings reported in this paper could be considered as a launching pad for extending the use of the gated silicon nanowire gas sensors for discriminations between polar and nonpolar analytes in complex, real-world gas mixtures. KEYWORDS: silicon, nanowire, transistor, sensor, hysteresis
■
INTRODUCTION Gated silicon nanowire (Si NW) devices, such as Si NW field effect transistors (FETs), have emerged as promising devices for chemical and biological sensing applications.1−8 In a standard back gate Si NW FET configuration, the electronic transport through the Si NW is affected by the periphery surfaces, interfaces and/or adsorbed atmosphere molecules near the charge carrier channel.9−15 Generally, these interactive effects lead to a “hysteresis” phenomenon, namely, a lag in the response obtained in the forward and backward electrical scans of the source-drain current (Ids) vs back-gate voltage (Vbgs).16−23 The hysteresis effect appears because applying a large gate bias lead to the injection of charge from (into) the Si NW into (from) the Si NW atop (oxide) surface sites, where the charge is trapped until the gate polarity is reversed.13,18,24 The higher is the environmental humidity or confounding factors, such as those appear in real-world conditions,25 the higher is the hysteresis and the higher is the screening of the targeted analytes. This essential drawback currently limits the widespread use of Si NW FETs in real-world environmental monitoring, homeland security, quality control in the food industry, medical sensing, and other areas (cf., refs 3 and 26). © 2012 American Chemical Society
It is believed that the Si NW FET hysteresis is caused by surface hydroxyl (Si−OH) sites,27−29 where water, Si−O¯, Si− OH2+, OH¯, or H+ species (hereby, surface trap states) originally exist.13,24,30 These trap states could be removed by thermal passivation,18,21,22,24 vacuum, or by chemical passivation via direct Si−C bond of nonoxidized Si NWs.30,31 A conceptually simpler and more cost-effective approach to remove these trap states and the associated hysteresis effect is based on the functionalization of the Si NW’s oxide sheath by a monolayer of trichlorosilane (TS) molecules via Si−O−Si bonds.7,32−36 In this context, it has been believed that the longer is the chain length of the TS molecules or the higher is the molecular density or hydrophobicity of the TS monolayers the lower is the hysteresis effect.37,38 Here, we show that the hydrophobicity and/or molecular density of the TS monolayers35,36 does not necessarily play a crucial role in determining the hysteresis in Si NW FET gas sensors. Rather, we show that the critical effect on the hysteresis of gas Si NW FET sensors2 is the concentration of the exposed or Received: February 19, 2012 Accepted: April 23, 2012 Published: April 23, 2012 2604
dx.doi.org/10.1021/am300288z | ACS Appl. Mater. Interfaces 2012, 4, 2604−2617
ACS Applied Materials & Interfaces
Research Article
after exposure to analytes. Source-drain current (Ids) versus voltage dependent back-gate (Vbgs) measurements, swept backward and forward between +40 V to −40 V with 200 mV steps and at 1 V source-drain voltage (Vds) and a sweep rate of 3.2 V/s, were carried out as follows:
unpassivated Si−OH groups (trap states) within the adsorbed TS monolayer. We reach this conclusion by a series of surface analysis and electrical measurements of TS molecules whose adsorption characteristics on Si NWs can be changed systematically.37,38 On the basis of these results, we provide a tentative understanding of (i) the relation between the adsorbed TS molecules, the hysteresis, or the fundamental parameters of FETs (including threshold voltage, carrier mobility, subthreshold swing, off-current, off-voltage) and of (ii) the relation between the hysteresis drift and possible screening effect on Si NW FET sensors upon exposure to different analytes at realworld conditions. The implications of the obtained results are presented and discussed in the text.
■
(i) after a sequential process of 1 min precleaning by di-ionized water, 5 min sonication with chloroform, and 12 h drying in 110 °C vacuum oven (hereafter, OHlow); (ii) after 2 min ultraviolet-ozone cleaning (UVOCs) that enriches the Si NW oxide surface with Si−OH groups (hereafter, OHhigh); (iii) after modification with controlled, TSAPR-made TS monolayer (hereafter, TSc) or after uncontrolled, direct self-assembly27 of TS monolayer (hereafter, TSuc); and (iv) upon exposure of the OHlow-, OHhigh-, TSc-, and TSuc-modified Si NW FETs to different concentrations of nonpolar and polar analytes (see Table 2) or humidity conditions. Each of the sequential four steps were first monitored under a flow of reference air (15% relative humidity (RH) and 99% purity; 0) leads to a positive shift of the Vth of the OHhigh-Si NW FET backward scan compared to the Vth of the OHlow-Si NW FET backward scan (ΔVth = +6.7 V, see Table 3). The large decrease in Qcss of the HTSc-Si NW FET compared to the Qcss of the OHhigh-Si NW FET (ΔQcss < 0), leads to a negative shift of the Vth of the HTSc-Si NW FET backward scan compared to the Vth of the OHhigh-Si NW FET backward scan (ΔVth = −7.5 V; see Table 3). According to the relationship of eq 15, ΔVH is given by the expression
where SS0 is the SS of the OHlow-Si NW FET backward scan. Our calculations show that a change of about +0.271 V in EbnwDnw decreased the SS of the OHhigh-Si NW FET by −3.7 V/decade in comparison to the SS of the OHlow-Si NW FET and that a change of about −0.273 V in EbnwDnw increased the SS of the HTSc-Si NW FET by +4.6 V/decade in comparison to the SS of the OHhigh-Si NW FET. This change in EbnwDnw is larger from the thermal voltage (kbT/q = 0.026 V), meaning significant changes in the band bending inside the Si NW. We have recently shown that the band bending inside the Si NW is also expressed in the Schottky barrier (Ψb) of the source and drain contacts.1 It was shown that the higher the band bending, the higher the Ψb and the lower the Ioff. To express the effect of Ioff (the Schottky thermionic emission current) quantitatively,57 the following relationship was considered: (cf., refs 57 and 58):
⎛ qΨ ⎞ Ioff ∝ exp⎜ − b ⎟ ⎝ k bT ⎠
qΔQ css
(15) 2612
dx.doi.org/10.1021/am300288z | ACS Appl. Mater. Interfaces 2012, 4, 2604−2617
ACS Applied Materials & Interfaces
Research Article
a point near the Voff, ΔΨb approaches zero, because the maximum band bending of the forward and backward scans is reached. For any higher positive Vbgs, (GF − GB)/GB is zero, with characteristic noise that originates from Ids of the order of
